bzr mirroring and KeyboardInterrupt

Asked by Daniel Fischer

Sometimes, when setting up a mirror for a bzr branch, mirroring fails with the following message (or similar):

   "Launchpad has not been able to mirror this branch. The last attempt was 56 minutes ago. (KeyboardInterrupt) Launchpad will try again shortly."

Repeatedly hitting "try again" every hour or so until the mirror is complete appears to fix the problem, but I'd rather know if there is something else that I'm supposed to do? Or is this just some sort of watchdog that kills a mirror attempt after ~half an hour?

(Btw., the message remains "will try again shortly" during the mirroring process, there is no indication that the process was already re-started.)

Question information

English Edit question
Launchpad itself Edit question
No assignee Edit question
Solved by:
Michael Hudson-Doyle
Last query:
Last reply:
Revision history for this message
Best Michael Hudson-Doyle (mwhudson) said :

The KeyboardInterrupt is a bit misleading. It means that the mirroring gave no indication of progress for some interval (15 minutes currently I think).

We detect progress by hooking into bzr's progress bar stuff and I think it's a bit of a known bug that branching pack branches can not update the progress bar even when the branching is progressing just fine, but I have to admit I thought we'd balanced the time out and bzr's behaviour about right by now. Apparently not.

That you can't see when a mirror attempt is in progress sounds like a fixable (and separate) problem -- can you file a bug for that?

Revision history for this message
Daniel Fischer (dannythefool) said :

Thanks Michael Hudson, that solved my question.

Revision history for this message
Daniel Fischer (dannythefool) said :

Thanks. These branches are fairly large for launchpad (close to 60k revisions), so it may be balanced after all :-)

Revision history for this message
Monty Taylor (mordred) said :

But is another good example of where this bug can be really problematic... when branching normally takes 20 to 30 minutes. :)

Revision history for this message
Tim Penhey (thumper) said :

The stacked branches work that is going on should make this *much* faster;-)